“How long will you need for multi-tenancy?” There are two bad answers to this question: “I don't know” and “six weeks.” The first helps nobody, the second sounds certain and isn't. This lesson shows a third: a range with a probability, assembled from parts, with the uncertainties that make it wide, and a plan to make it narrower. And it separates what architects often mix up: what we estimate, what the business wants and what we commit to.
A range instead of a number
An estimate is a statement about the future, and the future is uncertain. A single number hides that. “Six weeks” can mean: “six weeks at the earliest,” “probably six weeks” or “six weeks if nothing goes wrong,” and whoever hears the number usually picks the most favorable reading.
A range says more: “With 85% probability between 26 and 35 person-days, most likely around 31.” It shows three things at once: where you expect the work, how sure you are and where the risk lies. And making it honest is not a weakness. Whoever names a wide range says: here we still know too little, and that is information that management can react to.
Ranges get narrower over time. Before the first conversation with the customer, an estimate is often uncertain by a multiple; after a proof of concept and the first finished parts, the range shrinks. An estimate is therefore not a promise that you give once, but a status that you renew with every piece of new knowledge.